Curation Policy
PRISM is a curated resource hub for AI governance, release readiness, risk management, evaluation, and responsible AI deployment.
This file defines what belongs here, what does not, and how resources should be reviewed over time.
Inclusion Criteria
A resource should be included when it meets at least two of the following criteria:
- published by a recognized standards body, regulator, research institution, major open-source project, or credible practitioner organization
- directly useful for AI governance, release readiness, risk management, evaluation, observability, incident response, documentation, or compliance
- practical enough for a team to apply, not only read conceptually
- actively maintained, widely cited, or historically important
- includes clear documentation, public access, or reproducible artifacts
Exclusion Criteria
Do not add resources that are primarily:
- marketing pages with little reusable substance
- vendor claims without transparent methodology
- duplicated links already covered by a better primary source
- outdated summaries when an authoritative current version exists
- low-quality blog posts without evidence, examples, or practitioner value
Source Preference
Prefer primary and durable sources in this order:
- official standards and regulatory sources
- original research papers or project documentation
- reputable open-source repositories
- practitioner guides from credible organizations
- secondary explainers only when they add clarity not available in primary sources
Freshness Discipline
AI governance and evaluation resources change quickly. Use these review rules:
| Resource type | Review cadence |
|---|---|
| Regulations and standards | every 6 months |
| Active open-source tools | every 3 months |
| Benchmarks and eval frameworks | every 3 months |
| Academic papers | yearly unless superseded |
| Communities and courses | every 6 months |
When reviewing a section, check:
- whether the link still works
- whether the project is still maintained
- whether a newer version exists
- whether the description is still accurate
- whether the resource still belongs in the category
Description Standard
Each entry should explain why the resource matters in one sentence. Avoid vague descriptions such as “useful tool” or “good resource.”
Good:
Open-source LLM observability platform for tracing prompts, model outputs, latency, and cost across production applications.
Weak:
Useful observability tool.
Star Badges
GitHub star badges may be used for open-source repositories, but they should not replace judgment. A high-star project can still be out of scope, and a low-star project can still be valuable if it is technically strong or highly relevant.
Contribution Review Checklist
Before accepting a new resource, verify:
- the link works
- the source is credible
- the entry is not duplicative
- the description is specific and practical
- the category is correct
- the resource still fits the repository focus
Maintenance Notes
This repository should stay selective. A shorter, trusted resource list is more useful than a large directory of weak links.